ChatGPT has been helping me reconfigure my dad’s WordPress website (ricksegoine.net) so he can sell music downloads. Anyone who has ever tried to make WordPress do anything beyond being a basic blog understands the frustration: plugins that don’t talk to each other, settings that hide like they’re in witness protection, and error messages that might as well be written in Aramaic.
For the most part, ChatGPT walked me through it. It helped me troubleshoot, find the right plugins, and work around the parts of WordPress that only make sense at 2 a.m. with strong coffee. But then I hit the email setup, the place where even seasoned developers do the slow head shake.
No matter what combination of SMTP, host settings, and “try this other mysterious checkbox,” nothing worked. I seemed to be caught up in some sort of temporal causality loop, where I always ended up back at the same starting point (think Groundhog Day but with a website). Nothing I did seemed to work… until suddenly I somehow randomly escaped the loop by hitting just the right combination of checkboxes.
While ChatGPT did help shorten the path, it wasn’t able to reach into my website server and just fix it. That’s when I wondered: How far are we from having a real AI agent—Apple Intelligence, OpenAI, Grok, whoever—that can sit on my local machine, look at the problem, and just… solve it?
No explaining it
No guiding me through a difficult process
No guessing
It actually does the work
Are we already there and I’m missing something? Or am I still destined to keep chasing cryptic errors and clicking the same button twelve times in a row “just in case”? After a bit of research, I found that we are close, but not there quite yet.
The AI Co-Pilot Era (and Its Limits)
To be clear: we’ve come a long way. Even a couple years ago, trying to fix a WordPress email gateway issue would’ve meant:
- Googling random forum threads from 2014
- Watching out-of-date YouTube videos
- Reading documentation written by someone who definitely hates humans
With ChatGPT or Grok, I currently have access to an AI that can say, “Hey, your SMTP mail authentication might be failing because your host uses a different port. Try 587 instead of 465.” That’s already a huge win. But today’s AI is still a co-pilot. It assists, explains, warns, clarifies, and speeds things up, but it can’t take over the controls. Not yet at least. There are reasons for that:
- AI doesn’t yet have system-level access
- Security boundaries keep agents sandboxed
- Different apps store settings in different ways
- Companies don’t want AI touching sensitive configs without permission
So even though AI often sounds confident that it can fix something, it can’t just jump into my hosting service file manager and fix a half-deleted plugin folder. It can’t run migration scripts unless a human triggers them. It can’t read my server logs in real time unless I paste them in. But those limitations won’t last forever.
What Real AI Agents Will Look Like
This is where things get interesting. In the not-too-distant future, true AI agents—the kind I am waiting for—will be able to:
- Scan my local machine or website and figure out what’s installed
- Audit settings silently in the background
- Test configs, detect conflicts, and apply fixes
- Interact with apps directly (File Manager, cPanel, WordPress admin, plugins)
- Explain what they changed, and why
- Ask for permission when touching something risky
Imagine an AI agent that notices your SMTP email settings are failing because the YoastSEO WordPress plugin left a phantom directory inside the /wp-content/plugins folder. It then asks, “Want me to remove it and reinstall the latest version cleanly?” That’s when the game changes. AI is no longer a co-pilot, but has become a co-mechanic.
The 6–12 Month Window
We’re not quite there, but we’re closer than most people think. Every month, we edge a little further away from “language model” and a little closer to “autonomous assistant.” It is quite possible that within the next six to twelve months, we will have multimodal AI models that can see your screen, understand interfaces, interpret file structures, and manipulate tools. It’s not unrealistic to imagine that sometime soon, you’ll click a button and say: “Fix my WordPress email system.” And the agent just… does.
No outdated tutorials.
No port guessing.
No forums from 2008.
No more head-banging.
Until Then… It’s Wait (and Tinker)
For now, I’m still stuck clicking buttons. I still have to manually test settings. I’m still opening the file manager and muttering to myself like an old man yelling at a cloud.
But it’s way less painful than it was. And honestly? This “almost-there” phase is kind of fun too. It feels like being an early adopter again, watching the tools evolve in real time.
The future’s coming fast. And when the real AI assistants finally arrive, I’m looking forward to having far fewer bruises from banging my head against the tech wall.
Gabe Segoine
Gabe Segoine is the founder of LNKM and author of Surfing North Korea. He spent over a decade in Northeast Asia doing humanitarian work focused on the DPRK. Gabe now writes about culture, faith, and emerging AI from a deeply human perspective. His book Surfing North Korea and Other Stories from Inside can be found on Amazon.